sary before deciding whether the wave-length of the emitted light is greater than that of the exciting vibrations. My impression is that the bright narrow band in the yellow is slightly on the red side of the D lines, but I do not feel absolutely sure of it. It will be interesting to see whether the fluorescence persists for an appreciable time after the light is cut off, which can be easily determined with an especially designed phosphoroscope, a matter which I expect to take up next.
A quantitative study of the absorption of the vapour is extremely desirable, though the experimental difficulties will be very great. As I have already said, the width of the band at the D lines is often from ten to fifteen times the distance between the lines. Julius has, however, called attention to the fact that we must be on our guard against attributing the absence of light in the spectrum to absorp- tion, when the conditions are such that the rays may have been merely turned to one side by refraction. This he believes, if I read his paper rightly, is the true explanation of the widening of the D lines in the absorption spectrum. To eliminate the possibility of this lateral deviation it is necessary, if we are working with a non- homogeneous medium, to arrange matters so that the rays of light are perpendicular to the surfaces of equal refractive index, instead of parallel, as is the case in the dispersion tubes. One obvious method of getting around the difficulty would be to vaporise the sodium in a vacuum, thus doing away with the variations in the density, but this necessitates contact between the corrosive vapour and the glass. A better plan appeared to be to send the light through the stratified vapour in such a direction that no lateral deviation could result, a matter of some difficulty until the expedient occurred to me of using the surface of the molten metal as a reflector, the rays thus twice traversing the non-homogeneous medium in a direction perpen- dicular to the equi-indicial surfaces.
A photograph of the absorption spectrum of the vapour in the red and yellow region, obtained with a 10-foot concave grating, is repro- duced on Plate 3, fig. 12. The D lines were photographed just below the spectrum for comparison. The flutings in the red are well shown in this picture, which is but one of a large number that have been taken. Measurements of wave-length have not yet been made, as I feel confident of getting better plates in the near future. The fine dark lines can be found on both sides of the heavy band at the D lines, and I have found that they are present throughout the entire spectrum. An enlargement of a portion of fig. 12 is shown in fig. 13, which gives a better idea of the appearance of the fluted bands.
On increasing the density of the vapour the red end of the spectrum finally disappears, and by employing a very powerful light in combina- tion with the method above alluded to of reflecting the light from the
